Device and method for intelligently measuring leakage of embankment-penetrating building in freezing and thawing environment

Abstract

The invention belongs to the technical field of hydraulic engineering safety monitoring, and particularly relates to an intelligent measurement device and method for leakage of a dike-penetrating building in a freeze-thaw environment, and the method comprises the steps: selecting a plurality of monitoring points on the dike-penetrating building, and arranging a multi-dimensional sensor module to synchronously collect freeze-thaw cycle and multi-dimensional parameters related to the leakage of the dike-penetrating building; constructing a hybrid wireless transmission architecture of an external wireless transmission unit of the data collector; the multi-dimensional parameters are preprocessed; key characteristic parameters are calculated from the preprocessed multi-dimensional parameters, and real-time risk indexes can be quantified based on the key characteristic parameters; predicting a future change trend according to historical data of the key characteristic parameters; establishing a grading early warning mechanism based on the real-time risk index and the change trend of the key characteristic parameters; and identifying and positioning the crack of the dike-penetrating building in the freezing and thawing environment. According to the invention, the accuracy and the intelligent degree of leakage measurement of the dike-penetrating building in the freezing and thawing environment are effectively improved, and the labor cost is saved.

Description

Technical Field

[0001] This invention belongs to the field of water conservancy project safety monitoring technology, and in particular relates to an intelligent measurement device and method for leakage of structures crossing dikes under freeze-thaw conditions. Background Technology

[0002] Structures penetrating dikes are a crucial component of water conservancy projects, encompassing key facilities such as culverts, aqueducts, and inverted siphons. Their structural integrity and seepage prevention performance directly determine the core flood control and water retention functions of dikes, impacting the safety of people's lives and property and the stable development of the social economy within the basin. In the cold, high-altitude regions of northern my country and many seasonal freeze-thaw areas, these structures face severe natural challenges year-round. Freeze-thaw cycles act as "invisible eroders," continuously threatening the safety baseline of water conservancy projects.

[0003] When winter temperatures in cold regions plummet below freezing, the soil moisture at the junction of structures and the embankment rapidly freezes and expands, generating tremendous frost heave force. As temperatures rise in spring, the thawing of the frozen soil loosens the soil structure and reduces its strength, leading to thaw settlement. This repeated cycle of frost heave and thaw settlement easily tears tiny cracks at the junction of structures and the embankment. These initial cracks, often less than a millimeter wide, become channels for rainwater and river water to seep in. As moisture continues to seep in, the cracks widen further in subsequent freeze-thaw cycles—the expansion of water during freezing widens the cracks, and the thawing water carries sediment, creating cavities. This vicious cycle causes the cracks to extend from the surface to deeper layers, spreading from localized areas to the entire structure. If these hazards are not detected and addressed in time, they can ultimately lead to piping, soil erosion, and even major safety accidents such as embankment collapse and damage to structures, causing fatal blows to water conservancy projects and incalculable losses.

[0004] Currently, monitoring technologies for seepage in structures crossing levees under freeze-thaw conditions still have many shortcomings that urgently need to be addressed, making it difficult to meet the actual needs of engineering safety protection. Regarding monitoring indicators, existing devices are mostly limited to single-parameter acquisition, either monitoring only seepage pressure or recording only ambient temperature, completely ignoring the coupling mechanism between "temperature-stress-seepage" during freeze-thaw cycles. Under freeze-thaw conditions, temperature changes dominate the frost heave and thaw settlement of the soil, thereby altering the soil stress state. Stress changes, in turn, affect the development of seepage channels, and the seepage process, in turn, exacerbates the uneven distribution of the temperature field. This complex effect of multiple intertwined and mutually influential physical fields cannot be fully captured by a single indicator, making it even more difficult to identify the weak signals of early crack initiation and evolution. Often, by the time obvious anomalies appear in the monitoring data, the seepage hazard has already developed into a serious problem.

[0005] In terms of environmental adaptability, existing monitoring equipment generally has weak freeze-thaw resistance. In cold regions, extreme winter temperatures can reach below -30°C. The sensitive elements of conventional sensors are prone to malfunction due to icing, and the interfaces and insulation layers of transmission lines can be damaged by frost heave, leading to signal interruption or data distortion. More importantly, most monitoring equipment exposed on the surface of the dike or shallowly buried within it lacks targeted insulation and protection designs. In winter, condensation easily forms inside the equipment, and in summer, it faces rain and intense sunlight. Under these alternating extreme environmental conditions, the long-term stability and data reliability of the equipment are significantly compromised, often resulting in "data gaps in winter and data drift in summer," failing to provide continuous and effective data support for seepage monitoring.

[0006] The problems of low intelligence in monitoring and delayed risk response are also prominent. Currently, most water conservancy projects still rely on traditional methods for leakage monitoring, requiring personnel to periodically carry instruments to the site for inspections and single-point measurements. In cold regions, the dike surface is slippery and icy in winter, and mosquitoes and dense vegetation are rampant in summer. Manual inspections are not only inefficient but also pose high safety risks and are difficult to achieve all-weather, full-coverage monitoring, easily resulting in monitoring blind spots. More importantly, the existing monitoring system lacks the ability for automatic data analysis, real-time risk diagnosis, and intelligent early warning. The collected data needs to be manually sorted and judged one by one, often taking several hours or even days to complete risk identification, resulting in a serious lag in leakage risk response and missing the best opportunity to deal with emergencies.

[0007] Therefore, considering the unique characteristics of the freeze-thaw environment in cold regions and the urgent need for leakage monitoring of structures penetrating dikes, there is a pressing need to develop a leakage measurement device and method capable of adapting to harsh freeze-thaw conditions, achieving coordinated monitoring of multiple parameters such as temperature, stress, and seepage, and possessing intelligent data analysis, real-time leakage risk assessment, and automatic early warning functions. This technological breakthrough will not only fill the current gap in monitoring technology but also provide accurate and reliable technical support for the safe operation of water conservancy projects in cold regions, possessing extremely urgent practical significance and important engineering value.

[0008] Existing technology CN115356051A discloses a dam seepage monitoring system, method, and electronic equipment, including a detection module and a cloud platform. The detection module is used to monitor the resistivity of the dam profile in real time through buried electrodes. The cloud platform is used to convert the acquired dam profile resistivity into dam profile moisture content and issue an early warning when the dam profile moisture content meets preset warning conditions. By burying electrodes within the dam body to monitor the dam profile resistivity in real time, controlling the burial depth and electrode spacing improves the real-time performance and accuracy of the detection data. After uploading the collected data to the cloud platform to convert the dam profile resistivity into dam profile moisture content, the dam seepage is detected and an early warning is issued based on the dam profile moisture content. The seepage measurement method is simple and easy to operate, and can conveniently realize dam seepage measurement. However, this invention only considers the single parameter of dam profile resistivity and cannot effectively capture the multi-physics coupling effect and early crack evolution signal during the freeze-thaw process, resulting in low measurement accuracy. Existing technology CN213867663U discloses a drainage and seepage measurement system for an earth-rock dam, including a perimeter intercepting drainage ditch and a slope drainage ditch connected to the perimeter intercepting drainage ditch; a T-shaped dam body seepage drainage ditch with a concrete cover is provided between the perimeter intercepting drainage ditch and the slope drainage ditch; the T-shaped dam body seepage drainage ditch is equipped with double rows of drainage pipes and buried drainage blind materials. This novel system effectively combines dam body drainage, slope drainage, and dam body seepage measurement facilities in an orderly manner, ensuring dam safety while reducing project costs to some extent; by setting up the T-shaped dam body drainage ditch and drainage pipes, the seepage drainage effect of the dam body can be effectively improved; a thin-walled measuring weir with a cover is cleverly set up after the T-shaped dam body drainage ditch, which can accurately monitor the dam body seepage. However, this method still uses primitive mechanical mechanisms and manual observation for seepage measurement, resulting in high measurement costs, low measurement accuracy, and a large degree of human variation in measurement. Summary of the Invention

[0009] To address the aforementioned problems, this invention proposes an intelligent measurement device and method for leakage in structures crossing levees under freeze-thaw conditions. Through an intelligent design encompassing the entire process of "monitoring—transmission—analysis," it achieves accurate monitoring and risk warning of leakage in structures crossing levees under freeze-thaw conditions. To realize the above solution, this invention provides the following technical solution: A method for intelligently measuring leakage of structures penetrating dikes under freeze-thaw conditions includes the following steps: S1: Select several monitoring points on the structure through the dike and deploy multi-dimensional sensor modules to synchronously collect multi-dimensional parameters related to freeze-thaw cycles and leakage of the structure through the dike. S2: Construct a hybrid wireless transmission architecture for the external wireless transmission unit of the data acquisition unit, which is used for the synchronous acquisition, local storage and remote transmission of multi-dimensional sensor data as described in step S1. S3: Preprocess the multi-dimensional parameters acquired, stored and transmitted in steps S1 to S2, including removing outliers and smoothing non-stationary signals; then standardize the processed multi-dimensional parameters. S4: Calculate key feature parameters from the multidimensional parameters preprocessed in step S3, and quantify the real-time risk index based on the key feature parameters. R Construct an LSTM neural network model to predict future trends based on historical data of key feature parameters; S5: Real-time risk index obtained from step S4 R A four-level hierarchical early warning mechanism was established based on the changing trends of key feature parameters output by the LSTM neural network model. S6: Construct a 3D convolutional neural network model and a 3D U-Net neural network model, and combine them with the DBSCAN clustering algorithm to identify and locate cracks in structures crossing embankments under freeze-thaw conditions.

[0010] Further, step S1 includes: S10: Determine the sensor deployment scheme based on engineering geological data, structural diagrams of structures crossing the embankment, and on-site survey results; during deployment, increase the number of sensors in key areas such as the water-facing side, the backwater side, and the lateral junction of the structures crossing the embankment. S11: Set up soil temperature and humidity sensors to collect the temperature and volumetric moisture content of the soil around the building in real time, in order to identify the freeze-thaw phase and moisture redistribution; S12: Use fiber optic displacement sensors to monitor the frost heave displacement of the building structure and the surrounding soil in real time, and analyze the potential damage state of the building. S13: By using a vibrating wire pore water pressure sensor, monitor the changes in seepage pressure in the soil surrounding the building during freeze-thaw cycles and identify the intensity of leakage.

[0011] S14: Sensor installation involves burying the sensor in the designated location by excavating trenches and drilling holes, and then backfilling and compacting the soil.

[0012] Further, step S3 includes: S30: Eliminate outlier values ​​in multidimensional parameters using the 3σ criterion; S31: A window-adjustable exponentially weighted moving average algorithm is used to smooth high-frequency fluctuating non-stationary data, including parameters such as temperature and moisture content. S32: A resampling method based on linear interpolation is used to unify the data streams collected by each sensor to a standard time axis.

[0013] Further, step S4 includes: S40: Four key characteristic parameters for leakage determination are calculated from the multidimensional parameters after preprocessing in step S3, including temperature gradient, water content change rate, seepage pressure gradient and displacement rate. S41: A quantifiable real-time risk index is obtained based on four key feature parameters using the fuzzy comprehensive evaluation method. R ; S42: Construct an LSTM neural network model, with a data acquisition time of [time value missing]. n The historical data values ​​of the four key feature parameters were used to divide the dataset into training, testing, and validation datasets. n These are variables that can be determined independently; S43: Train the LSTM neural network model on the training and validation datasets, using the mean squared error loss function and saving the weight parameters; perform performance testing on the test dataset, select the LSTM neural network model with good test performance for deployment, and output the predicted values ​​of the four key feature parameters. S44: Based on the predicted values ​​of the four key feature parameters obtained in step S43, the predicted value of the risk index is obtained using the same method as in step S41. .

[0014] Furthermore, the LSTM neural network model described in step S42 adopts a two-layer LSTM unit structure, with Dropout operation introduced between layers, ReLU activation function used in the hidden layer, and linear activation function used in the output layer.

[0015] Further, step S5 includes: R < L 1. Under normal conditions, all parameters are within the safe threshold, and only data is recorded; L 2≤ R < L 3. Monitor status; if individual parameters show abnormalities, mark the abnormal values ​​to alert the user. L 3≤ R < L 4. Warning status: When multiple parameters show joint anomalies and the risk of LSTM prediction increases, warning SMS and emails are sent via pre-set phone numbers and email addresses. R ≥ L 5. Alarm status: If the risk index exceeds the limit or the parameter is seriously abnormal, an on-site audible and visual alarm will be triggered and an emergency message will be pushed. in, L 1. L 2. L 3. L 4. LAll 5 are variable values ​​that can be determined independently based on actual circumstances, and satisfy the following conditions: L 1< L 2< L 3< L 4< L 5.

[0016] Further, step S6 includes: S60: Early warning mechanism and real-time risk index constructed based on step S5 R The monitoring points that have entered the warning state are selected, which are the monitoring points that may have cracks; S61: The monitoring data of the monitoring points selected in step S60 are interpolated by introducing the Kriging interpolation method, and the monitoring point data are estimated by the optimal linear unbiased estimation, which is then transformed into a three-dimensional feature map. S62: For the three-dimensional feature map generated in step S61, a three-layer 3D convolutional neural network is used to perform multi-scale feature extraction to obtain multi-channel features; S63: Construct a three-dimensional U-Net neural network model based on an encoder-decoder architecture, and use the multi-channel features obtained in step S62 as the encoder input to output the probability of crack existence and crack orientation information at the monitoring point. S64: Settings m A preliminary crack presence probability binarization is performed using a threshold as the initial threshold. Voxels with probabilities higher than the threshold are initially identified as crack regions, resulting in a preliminary crack voxel set. m These are variables that can be determined independently; S65: Apply noisy robust DBSCAN clustering to the voxel set to obtain continuous crack clusters, and perform geometric reconstruction on the continuous crack clusters. Finally, output the crack existence probability voxel map, the spatial coordinates and volume parameters of the continuous crack clusters, crack direction angle data, and a three-dimensional crack network model.

[0017] Furthermore, the 3D convolutional neural network model uses 3×3×3 convolutional kernels and Leaky ReLU activation function in the shallow layers; 5×5×5 convolutional kernels and ReLU activation function in the middle layers; and 7×7×7 convolutional kernels and Swish activation function in the deep layers.

[0018] Furthermore, the encoder of the 3D U-Net neural network model reuses and fine-tunes the pre-trained weights of the 3D convolutional neural network model; the decoder gradually upsamples through 4 layers of 3×3×3 transposed convolutional kernels, and each layer of the decoder fuses the feature maps of the encoder at the corresponding scale through skip connections; the output layer uses a 1×1×1 convolutional kernel to output two independent channels, which correspond to the probability of crack existence and crack direction information, respectively.

[0019] On the other hand, an intelligent measurement device for leakage of structures crossing dikes under freeze-thaw conditions is provided, which is applied to any of the intelligent measurement methods for leakage of structures crossing dikes under freeze-thaw conditions as described in any one of the claims. The device includes a soil moisture sensor 1, a displacement sensor 2, a pore water pressure sensor 3, a data acquisition and protection box 6, an intelligent power supply controller 7, and a hybrid power supply module. Soil moisture sensor 1, displacement sensor 2, and pore water pressure sensor 3 are each connected to an independent cold-resistant polyethylene sheathed cable 4. The rear half of the three cold-resistant polyethylene sheathed cables 4 are uniformly wrapped by a PVC pipe 5 and connected to the data acquisition protective box 6. The data acquisition and protection box 6 is connected to an external intelligent power controller 7, and the intelligent power controller 7 is connected to the solar panel 9 and the storage battery respectively through the copper core PVC sheathed flexible wire 10. The hybrid power supply module is powered by a combination of solar panel 9 and battery 11. The battery 11 is wrapped with a rock wool insulation layer 12, and the rock wool insulation layer 12 is covered with a shell made of corrosion-resistant and damage-resistant glass fiber reinforced plastic 13. A solar panel bracket 8 is installed at the bottom of the solar panel 9. The soil moisture sensor 1 has three sensor probes installed at the front end. The sensor is covered with a polytetrafluoroethylene composite material protective kit 15 and is connected to a cold-resistant polyethylene sheathed cable 4 at the rear output port. The connection is provided with a nitrile rubber sealing ring 16. The intelligent power supply controller 7 has a display screen 20, three indicator lights 19 and an on / off button 17 installed on its panel. A wireless transmission antenna 18 is installed on the right side. A nitrile rubber sealing ring 16 is provided at the connection between the front end of the intelligent power supply controller 7 and the copper core PVC sheathed flexible wire 10. The data acquisition protective box 6 has an outer shell of ABS engineering plastic 21, filled with a thick rock wool insulation layer 22. A loudspeaker 31 is embedded in the front side, and a rain shield 32 is installed on top of the loudspeaker 31. Two identical heating elements 23 are installed on the rear side of the interior, with the width of each element significantly smaller than the width of the rear surface. A power supply interface 24 is located between the two heating elements 23. A DTU wireless data transmission module 28 is attached to the heating element 23 on the right side of the rear side. A wireless transmission module display screen 35 and an on / off button 29 are installed on the upper surface of this module. A wireless transmission antenna 34 is installed on the right side, its direction parallel to the side of the data acquisition protective box 6. The front surface of the DTU wireless data transmission module 28 is connected to a wireless transmission interface 27, with an RS-232 interface 26 installed on top of the wireless transmission interface 27. An LED bulb 33 is installed in the lower right corner of the interior of the data acquisition protective box 6, connected to the bottom surface of the interior of the data acquisition protective box 6 via a bulb base 30. A data acquisition interface 25 is installed on the left side of the interior of the data acquisition protective box 6.

[0020] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention achieves in-depth analysis of multi-field coupling effects by simultaneously collecting multi-dimensional parameters such as osmotic pressure, temperature, moisture, and displacement, thus avoiding the limitations of single-parameter monitoring.

[0021] 2. The protective design of the thermal insulation module effectively solves the problem of easy failure of sensors, cables and data acquisition equipment in harsh freeze-thaw environments, ensuring the continuity of long-term monitoring. Attached Figure Description

[0022] Figure 1 A schematic diagram of the overall scheme for an intelligent measurement device and method for leakage of structures crossing dikes under freeze-thaw conditions.

[0023] Figure 2 Feature extraction and risk index for key feature parameters R Flowchart for future value prediction.

[0024] Figure 3 This is a flowchart for crack identification and location.

[0025] Figure 4 This is a structural diagram of an intelligent measurement device for seepage in structures crossing dikes under freeze-thaw conditions.

[0026] Figure 5 Detailed diagram of sensor protection.

[0027] Figure 6 Detailed diagram of the intelligent power supply controller.

[0028] Figure 7 This is a diagram of the internal structure of the data acquisition and protection box. Detailed Implementation

[0029] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0030] In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0031] In the description of this application, the term "for example" is used to mean "used as an example, illustration, or description." Any embodiment described as "for example" in this application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use the invention. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that the invention can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid obscuring the description of the invention with unnecessary detail. Therefore, the invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed in this application.

[0032] In this embodiment of the invention, an intelligent method for measuring leakage of structures penetrating dikes under freeze-thaw conditions is described, as detailed in the appendix. Figure 1 ~Appendix Figure 3 Specifically, it includes: S1: Select several monitoring points on the structure through the dike and deploy multi-dimensional sensor modules to synchronously collect multi-dimensional parameters related to freeze-thaw cycles and leakage of the structure through the dike. S2: Construct a hybrid wireless transmission architecture for the external wireless transmission unit of the data acquisition unit, which is used for the synchronous acquisition, local storage and remote transmission of multi-dimensional sensor data as described in step S1. S3: Preprocess the multi-dimensional parameters acquired, stored and transmitted in steps S1 to S2, including removing outliers and smoothing non-stationary signals; then standardize the processed multi-dimensional parameters. S4: Calculate key feature parameters from the multidimensional parameters preprocessed in step S3, and quantify the real-time risk index based on the key feature parameters. R Construct an LSTM neural network model to predict future trends based on historical data of key feature parameters; S5: Real-time risk index obtained from step S4 R A four-level hierarchical early warning mechanism was established based on the changing trends of key feature parameters output by the LSTM neural network model. S6: Construct a 3D convolutional neural network model and a 3D U-Net neural network model, and combine them with the DBSCAN clustering algorithm to identify and locate cracks in structures crossing embankments under freeze-thaw conditions.

[0033] See attached document Figure 4 ~Appendix Figure 7 The present invention also provides an intelligent measurement device for leakage of structures crossing dikes under freeze-thaw conditions, including a soil moisture sensor 1, a displacement sensor 2, a pore water pressure sensor 3, a data acquisition and protection box 6, an intelligent power supply controller 7, and a hybrid power supply module. Soil moisture sensor 1, displacement sensor 2, and pore water pressure sensor 3 are each connected to an independent cold-resistant polyethylene sheathed cable 4. The rear half of the three cold-resistant polyethylene sheathed cables 4 are uniformly wrapped by a PVC pipe 5 and connected to the data acquisition protective box 6. The data acquisition and protection box 6 is connected to an external intelligent power controller 7, and the intelligent power controller 7 is connected to the solar panel 9 and the storage battery respectively through the copper core PVC sheathed flexible wire 10. The hybrid power supply module is powered by a combination of solar panel 9 and battery 11. The battery 11 is wrapped with a rock wool insulation layer 12, and the rock wool insulation layer 12 is covered with a shell made of corrosion-resistant and damage-resistant glass fiber reinforced plastic 13. A solar panel bracket 8 is installed at the bottom of the solar panel 9. The soil moisture sensor 1 has three sensor probes installed at the front end. The sensor is covered with a polytetrafluoroethylene composite material protective kit 15 and is connected to a cold-resistant polyethylene sheathed cable 4 at the rear output port. The connection is provided with a nitrile rubber sealing ring 16. The intelligent power supply controller 7 has a display screen 20, three indicator lights 19 and an on / off button 17 installed on its panel. A wireless transmission antenna 18 is installed on the right side. A nitrile rubber sealing ring 16 is provided at the connection between the front end of the intelligent power supply controller 7 and the copper core PVC sheathed flexible wire 10. The data acquisition protective box 6 has an outer shell of ABS engineering plastic 21, filled with a thick rock wool insulation layer 22. A loudspeaker 31 is embedded in the front side, and a rain shield 32 is installed on top of the loudspeaker 31. Two identical heating elements 23 are installed on the rear side of the interior, with the width of each element significantly smaller than the width of the rear surface. A power supply interface 24 is located between the two heating elements 23. A DTU wireless data transmission module 28 is attached to the heating element 23 on the right side of the rear side. A wireless transmission module display screen 35 and an on / off button 29 are installed on the upper surface of this module. A wireless transmission antenna 34 is installed on the right side, its direction parallel to the side of the data acquisition protective box 6. The front surface of the DTU wireless data transmission module 28 is connected to a wireless transmission interface 27, with an RS-232 interface 26 installed on top of the wireless transmission interface 27. An LED bulb 33 is installed in the lower right corner of the interior of the data acquisition protective box 6, connected to the bottom surface of the interior of the data acquisition protective box 6 via a bulb base 30. A data acquisition interface 25 is installed on the left side of the interior of the data acquisition protective box 6.

[0034] Specifically, to address the problems of existing methods for measuring leakage in structures crossing levees under freeze-thaw conditions, such as limited parameter sources, low levels of measurement intelligence, low accuracy, high costs of manual inspections, and delayed risk response, safety engineers for these structures have adopted an intelligent method for measuring leakage in freeze-thaw environments. First, several monitoring points are selected on the levee structure, and multi-dimensional sensor modules are deployed to synchronously collect multi-dimensional parameters related to freeze-thaw cycles and leakage. A hybrid wireless transmission architecture is constructed, connecting the data acquisition unit to an external wireless transmission unit, for synchronous acquisition, local storage, and remote transmission of multi-dimensional sensor data. Then, the acquired, stored, and transmitted multi-dimensional parameters are preprocessed, including outlier removal and smoothing of non-stationary signals. The processed multi-dimensional parameters are then standardized. Key feature parameters are calculated from the preprocessed multi-dimensional parameters, and a quantifiable real-time risk index is generated based on these key feature parameters. R ; Construct an LSTM neural network model to predict future trends based on historical data of key feature parameters; Furthermore, based on the acquired real-time risk index R A four-level hierarchical early warning mechanism was established based on the changing trends of key feature parameters output by the LSTM neural network model. Finally, a 3D convolutional neural network model and a three-dimensional U-Net neural network model were constructed, and the DBSCAN clustering algorithm was combined to identify and locate cracks in structures crossing levees under freeze-thaw conditions. Simultaneously, an intelligent measurement device for leakage in structures crossing levees under freeze-thaw conditions was constructed, including a soil moisture sensor, a displacement sensor, a pore water pressure sensor, a data acquisition and protection box, an intelligent power supply controller, and a hybrid power supply module.

[0035] Example 1 In one embodiment, as shown in the appendix Figure 1 Appendix Figure 4 and attached Figure 5 As shown, to address the issue of limited parameter sources in current methods for measuring seepage in structures crossing levees under freeze-thaw conditions, safety engineers for these structures utilize multi-dimensional sensors to simultaneously collect multi-dimensional parameters, specifically including: S10: Determine the sensor deployment scheme based on engineering geological data, structural diagrams of structures crossing the embankment, and on-site survey results; during deployment, increase the number of sensors in key areas such as the water-facing side, the backwater side, and the lateral junctions of the structures crossing the embankment. S11: Set up soil temperature and humidity sensors to collect the temperature and volumetric moisture content of the soil around the building in real time, in order to identify the freeze-thaw phase and moisture redistribution; S12: Use fiber optic displacement sensors to monitor the frost heave displacement of the building structure and the surrounding soil in real time, and analyze the potential damage state of the building. S13: By using a vibrating wire pore water pressure sensor, monitor the changes in seepage pressure in the soil surrounding the building during freeze-thaw cycles and identify the intensity of leakage.

[0036] S14: Sensor installation involves excavating trenches and drilling holes to bury the sensors in the designated locations, and then backfilling and compacting the soil to ensure close contact between the sensors and the surrounding soil. S15: Remotely configure the sampling frequency of the data acquisition device via the cloud platform: temperature and moisture 10 min / time, osmotic pressure and displacement 5 min / time, external transmission module upload frequency 30 min / time, trial run for 48 hours to verify data consistency, and finally enter formal monitoring mode after stabilization.

[0037] Example 2 In one embodiment, as shown in the appendix Figure 1 Appendix Figure 4 As shown, to address the issues of limited parameter sources and delayed risk response in current methods for measuring seepage in structures crossing levees under freeze-thaw conditions, safety engineers for these structures have developed a hybrid wireless transmission architecture with an external wireless transmission unit for data acquisition. This architecture specifically includes: This system employs a combined architecture of a CR1000 data acquisition unit and an external wireless transmission unit to achieve simultaneous multi-sensor data acquisition, local storage, and remote transmission, adapting to the low-power and high-stability requirements of freeze-thaw environments. The CR1000 data acquisition unit boasts a wide operating temperature range of -25℃ to 50℃ and supports signal input from various sensor types. It features built-in 64MB of non-volatile storage, capable of storing ≥1 million monitoring data entries, and supports local data storage and resume transmission from where it stops, preventing data loss due to transmission interruptions. The wireless transmission unit communicates bidirectionally with the CR1000 data acquisition unit via a DTU wireless data transmission module and connects to the networks of the three major telecom operators, enabling long-distance remote data upload.

[0038] Example 3 In one embodiment, as shown in the appendix Figure 1 As shown, the safety engineer for the structure crossing the embankment preprocesses the collected multi-dimensional sensor data, specifically including: S30: Outlier values ​​in multidimensional parameters are removed using the 3σ criterion; the calculation expression is as follows:

[0039] in, For input data, The average value of the input data sequence. For standard deviation, calculate results exceeding the standard deviation. The data was considered outliers and removed. S31: The transmitted data is mostly non-stationary signals. An adjustable window exponentially weighted moving average algorithm is used to smooth the high-frequency fluctuating non-stationary data, including parameters such as temperature and moisture content, to effectively suppress random noise while preserving its true trend. The calculation formula is as follows:

[0040] in, for t The smoothing value at time, for t The original value at time, for t The smoothed value at time -1 This is a smoothing factor with a value range of (0,1), which determines the rate of weight decay. S32: A resampling method based on linear interpolation is used to unify the data streams collected by each sensor to a standard time axis.

[0041] Example 4 In one embodiment, as shown in the appendix Figure 1 Appendix Figure 2 As shown, to address the problems of low measurement intelligence, low measurement accuracy, and high manual inspection costs in existing methods for measuring leakage in structures crossing dikes under freeze-thaw conditions, safety engineers for structures crossing dikes have combined artificial neural network models to achieve intelligent and accurate prediction of leakage in these structures. Specifically, this includes: S40: Four key feature parameters for leakage determination are calculated from the multidimensional parameters after preprocessing in step S3, including the temperature gradient ( ), rate of change in water content ( ), osmotic pressure gradient ( ) and displacement rate ( ); S41: A quantifiable real-time risk index is obtained based on four key feature parameters using the fuzzy comprehensive evaluation method. R The fuzzy comprehensive evaluation model consists of three elements. U , V, H Composition, factor set for , Four evaluation factors, evaluation set ,in , , , These are respectively: safety, attention, early warning, and alarm; evaluation matrix. H The expression is as follows:

[0042] in, To evaluate the target through elements correspond The membership degree of the levels. Then, the factor weight set is determined using the analytic hierarchy process. In the formula , , , The values ​​are 0.2, 0.3, 0.35, and 0.15 respectively. Final Risk Index Output a comprehensive leakage risk index between 0 and 1. R ; S42: Constructing an LSTM neural network model for a comprehensive leakage risk index R Trend prediction, the LSTM calculation process is expressed as follows:

[0043]

[0044] in, and These are the input gate and the corresponding weight matrix, respectively. and These are the forget gate and the corresponding weight matrix, respectively. and These represent the output gate and the corresponding weight matrix, respectively. The state matrix, Represents the candidate matrix, tanh is the activation function. Enter the current time. The output from the previous moment. Indicates the output at the current time; The LSTM neural network model employs a two-layer LSTM unit structure, with each layer containing 64 neurons. Dropout (ratio 0.2) is introduced between layers to prevent overfitting. ReLU is used as the activation function in the hidden layers, and a linear activation function is used in the output layer. Model training utilizes batch gradient descent with a batch size of 64, a learning rate of 0.001, 200 iterations, and a time step of 10. The mean squared error (MSE) is chosen as the loss function, calculated using the following formula:

[0045] in, For the true value, For predicted values, n The sample size is given. The model uses historical data of four core feature parameters as the training set to predict the trend of feature parameter changes in the next 24 hours, and further derives the trend of leakage risk index changes. Historical data values ​​of four key feature parameters were collected over a period of 96 hours, and the training dataset, test dataset, and validation dataset were divided in a 7:2:1 ratio. S43: Train the LSTM neural network model on the training and validation datasets, using the mean squared error loss function and saving the weight parameters; perform performance testing on the test dataset, select the LSTM neural network model with good test performance for deployment, and output the predicted values ​​of the four key feature parameters. S44: Based on the predicted values ​​of the four key feature parameters obtained in step S43, the predicted value of the risk index is obtained using the same method as in step S41. .

[0046] Example 5 In one embodiment, as shown in the appendix Figure 1 As shown, to address the issue of delayed risk response in existing methods for measuring leakage in structures penetrating dikes under freeze-thaw conditions, safety engineers for these structures use real-time risk indices as a basis. R A four-level hierarchical early warning mechanism was established based on the changing trends of key feature parameters output by the LSTM neural network model, specifically including: R <0.3, normal state, all parameters are within the safe threshold, the cloud platform only records data; 0.3≤ R <0.6, under watch status, individual parameters are abnormal, the cloud platform marks the abnormal value to prompt attention; 0.6≤ R <0.8, warning status, multiple parameters are abnormal and the risk of LSTM prediction is increased, warning SMS and email will be sent through the phone number and email address recorded by the platform; R ≥0.8 indicates an alarm state, where the risk index exceeds the limit or the parameter is severely abnormal, triggering an on-site audible and visual alarm and pushing emergency information.

[0047] Example 6 In one embodiment, as shown in the appendix Figure 1 Appendix Figure 3 As shown, safety engineers for structures penetrating embankments use convolutional neural network models to implement early warning risk indices. R Accurate crack identification and location are achieved when the crack strength is ≥0.6, specifically including: S60: Early warning mechanism and real-time risk index constructed based on step S5 R Filter out R Monitoring points with a value ≥0.6 are those where cracks may exist; S61: Due to the uneven distribution of monitoring nodes, interpolation errors can occur. Therefore, Kriging interpolation is introduced to interpolate the monitoring data of the monitoring points selected in step S60, and optimal linear unbiased estimation is performed on the monitoring point data, which is then transformed into a three-dimensional feature map. The unbiased estimate is expressed as a weighted sum of the monitoring point data, and the calculation expression is:

[0048] in, For target point The estimated value, N For the number of monitoring points, For the first i The weighting coefficient of each monitoring point For the first i Measured data from each monitoring point; Weight Obtained by solving a system of linear equations:

[0049] in, C The covariance matrix between sample points This is the weight coefficient vector. c To monitor sample points and target points The covariance vector between them The Lagrange multiplier is used. Through the above steps, the monitoring data is interpolated into a regular three-dimensional grid field, generating a three-dimensional feature map with a resolution of 0.1 meters and a size of 128×128×128. S62: For the 3D feature map generated in step S61, a three-layer 3D convolutional neural network (3D-CNN) is used for multi-scale feature extraction to obtain multi-channel features. The 3D convolution is performed in spatial position ( x , y , z ) and channel n The calculation on can be expressed as:

[0050] in, To output the feature map in spatial location ( x , y , z ) n The values ​​of each feature channel; I For inputting a three-dimensional tensor; Let n be the weight matrix of the nth 3D convolution kernel, with the shape ( K , K , K , C in ); For the first n The bias terms corresponding to each convolution kernel; Input the number of channels; K The kernel size; To capture crack features at different scales, the 3D-CNN network structure consists of a shallow layer using 3×3×3 convolutional kernels and a Leaky ReLU activation function to capture local parameter abrupt changes, corresponding to the initial crack formation region; a middle layer using 5×5×5 convolutional kernels and a ReLU activation function to extract parameter gradient features along the crack extension direction, corresponding to the crack extension direction; and a deep layer using 7×7×7 convolutional kernels and a Swish activation function to identify multi-region parameter co-variation features, corresponding to the crack network continuity state; finally, a 256-channel 3D feature map is output.

[0051] S63: Using the multi-channel features extracted by 3D-CNN as encoder input, a 3D U-Net based on an encoder-decoder architecture is constructed to invert the feature maps into an interpretable crack parameter field. The encoder part reuses and fine-tunes the pre-trained weights of the aforementioned 3D-CNN. The decoder progressively upsamples through four layers of 3×3×3 transposed convolutional kernels. Each decoder layer fuses the feature maps of the corresponding scale from the encoder through skip connections. The output layer uses a 1×1×1 convolutional kernel to output two independent channels, corresponding to the crack existence probability and crack direction information, respectively. S64: A threshold of 0.5 is set for initial binary judgment of crack presence probability. Voxels with probabilities higher than the threshold are initially identified as crack regions, resulting in a preliminary crack voxel set. To eliminate isolated noise and identify continuous crack clusters, a noisy robust DBSCAN clustering method is applied to the voxel set, with a neighborhood radius of 0.2 meters and a minimum number of points of 5. Within the crack volume obtained from each cluster, the local direction vector is calculated based on the strain difference of adjacent voxels on the three axes, and the strike angle is calculated after projecting it onto the horizontal plane. :

[0052] in, and They are respectively X and Y Strain difference component in the direction.

[0053] S65: Geometric reconstruction of continuous crack clusters obtained by DBSCAN clustering. Input the voxel coordinates and attribute data of the crack clusters, calculate the isosurfaces between voxels through interpolation, convert discrete voxels into continuous surface meshes, and generate a quantifiable 3D crack network model. The final output includes a crack existence probability voxel map, spatial coordinates and volume parameters of the continuous crack clusters, crack strike angle data, and the 3D crack network model.

[0054] Example 3 In one embodiment, as shown in the appendix Figure 1 Appendix Figure 4 - Appendix Figure 7 As shown, a safety engineer for structures crossing levees has constructed an intelligent leakage measurement device for structures crossing levees under freeze-thaw conditions. This device is applied to any of the intelligent leakage measurement methods for structures crossing levees under freeze-thaw conditions described in the invention. The device includes a soil moisture sensor 1, a displacement sensor 2, a pore water pressure sensor 3, a data acquisition and protection box 6, an intelligent power supply controller 7, and a hybrid power supply module. Specifically, it includes: Soil moisture sensor 1, displacement sensor 2, and pore water pressure sensor 3 are each connected to an independent cold-resistant polyethylene sheathed cable 4 to provide freeze-thaw protection for the equipment and prevent component failure due to freeze-thaw. The thickness of the cold-resistant polyethylene sheathed cable 4 is 5-8 mm. The rear half of the three cold-resistant polyethylene sheathed cables 4 are uniformly wrapped by a PVC pipe 5 and connected to the data acquisition protective box 6. The data acquisition and protection box 6 is connected to an external intelligent power controller 7, and the intelligent power controller 7 is connected to the solar panel 9 and the storage battery respectively through the copper core PVC sheathed flexible wire 10. The hybrid power supply module is powered by a combination of solar panels 9 and batteries 11. Batteries 11 use two 12V / 100Ah gel batteries with lead-calcium alloy plates. The operating temperature range is -30℃ to 60℃. Batteries 11 are wrapped with a 50mm thick rock wool insulation layer 12, and the rock wool insulation layer 12 is covered with a shell made of corrosion-resistant and damage-resistant glass fiber reinforced plastic 13. Solar panels 9 use 300W monocrystalline silicon solar panels with dimensions of 1650mm×990mm×35mm. The operating temperature range is -40℃ to 85℃, the installation tilt angle is set to 45°, and a solar panel bracket 8 is installed at the bottom. The soil moisture sensor 1 has three sensor probes installed at the front end. The sensor is covered with a polytetrafluoroethylene composite material protective kit 15 and is connected to a cold-resistant polyethylene sheathed cable 4 at the rear output port. The connection is provided with a nitrile rubber sealing ring 16. The intelligent power supply controller 7 has a display screen 20, three indicator lights 19 and an on / off button 17 installed on its panel. A wireless transmission antenna 18 is installed on the right side. A nitrile rubber sealing ring 16 is provided at the connection between the front end of the intelligent power supply controller 7 and the copper core PVC sheathed flexible wire 10. The data acquisition protective box 6 has an outer shell of ABS engineering plastic 21, filled with a 25mm thick rock wool insulation layer 22. A loudspeaker 31 is embedded on the front side, with a rain shield 32 mounted on top of the loudspeaker 31. Two identical heating elements 23 are installed on the rear side of the interior, with the width of each element significantly smaller than the width of the rear surface. A power interface 24 is located between the two heating elements 23. A 10W heating element 23 is installed on the right side of the rear side of the interior, activating heating when the internal temperature is below -20℃. The heating elements 23 are closely attached to a DTU wireless data transmission module 28, on the upper surface of which is mounted a wireless transmission module display screen 35 and a... The power button 29 has a wireless transmission antenna 34 mounted on its right side, with its direction parallel to the side of the data acquisition protective box 6. The front surface of the DTU wireless data transmission module 28 is connected to the wireless transmission interface 27, and an RS-232 interface 26 is mounted on the upper part of the wireless transmission interface 27. An LED bulb 33 is installed in the lower right corner inside the data acquisition protective box 6, and is connected to the bottom surface inside the data acquisition protective box 6 through the bulb base 30. A data acquisition interface 25 is installed on the left side inside the data acquisition protective box 6. To prevent water from seeping into the components, nitrile rubber sealing rings are added to all interfaces, and waterproof strips are added to the door frame of the protective box.

[0055] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0056] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. An intelligent method for measuring leakage in structures crossing dikes under freeze-thaw conditions, characterized in that, Includes the following steps: S1: Select several monitoring points on the structure through the dike and deploy multi-dimensional sensor modules to synchronously collect multi-dimensional parameters related to freeze-thaw cycles and leakage of the structure through the dike. S2: Construct a hybrid wireless transmission architecture for the external wireless transmission unit of the data acquisition unit, which is used for the synchronous acquisition, local storage and remote transmission of multi-dimensional sensor data as described in step S1. S3: Preprocess the multi-dimensional parameters acquired, stored and transmitted in steps S1 to S2, including removing outliers and smoothing non-stationary signals; then standardize the processed multi-dimensional parameters. S4: Calculate key feature parameters from the multidimensional parameters preprocessed in step S3, and quantify the real-time risk index based on the key feature parameters. R Construct an LSTM neural network model to predict future trends based on historical data of key feature parameters; S5: Real-time risk index obtained from step S4 R A four-level hierarchical early warning mechanism was established based on the changing trends of key feature parameters output by the LSTM neural network model. S6: Construct a 3D convolutional neural network model and a 3D U-Net neural network model, and combine them with the DBSCAN clustering algorithm to identify and locate cracks in structures crossing embankments under freeze-thaw conditions.

2. The intelligent measurement method for seepage of structures crossing dikes under freeze-thaw conditions according to claim 1, characterized in that, Step S1 includes: S10: Determine the sensor deployment scheme based on engineering geological data, structural diagrams of structures crossing the embankment, and on-site survey results; during deployment, increase the number of sensors in key areas such as the water-facing side, the backwater side, and the lateral junction of the structures crossing the embankment. S11: Set up soil temperature and humidity sensors to collect the temperature and volumetric moisture content of the soil around the building in real time, in order to identify the freeze-thaw phase and moisture redistribution; S12: Use fiber optic displacement sensors to monitor the frost heave displacement of the building structure and the surrounding soil in real time, and analyze the potential damage state of the building. S13: By using a vibrating wire pore water pressure sensor, the changes in seepage pressure in the soil surrounding the building during freeze-thaw cycles are monitored to identify the intensity of leakage. S14: Sensor installation involves burying the sensor in the designated location by excavating trenches and drilling holes, and then backfilling and compacting the soil.

3. The intelligent measurement method for seepage of structures crossing dikes under freeze-thaw conditions according to claim 1, characterized in that, Step S3 includes: S30: Eliminate fluctuation outliers in multidimensional parameters using the 3σ criterion; S31: A window-adjustable exponentially weighted moving average algorithm is used to smooth high-frequency fluctuating non-stationary data, including parameters such as temperature and moisture content. S32: A resampling method based on linear interpolation is used to unify the data streams collected by each sensor to a standard time axis.

4. The intelligent measurement method for seepage of structures crossing dikes under freeze-thaw conditions according to claim 1, characterized in that, Step S4 includes: S40: Four key characteristic parameters for leakage determination are calculated from the multidimensional parameters after preprocessing in step S3, including temperature gradient, water content change rate, seepage pressure gradient and displacement rate. S41: A quantifiable real-time risk index is obtained based on four key feature parameters using the fuzzy comprehensive evaluation method. R ; S42: Construct an LSTM neural network model, with a data acquisition time of [time value missing]. n The historical data values ​​of the four key feature parameters were used to divide the dataset into training, testing, and validation datasets. n These are variables that can be determined independently; S43: Train the LSTM neural network model on the training and validation datasets, using the mean squared error loss function and saving the weight parameters; perform performance testing on the test dataset, select the LSTM neural network model with good test performance for deployment, and output the predicted values ​​of the four key feature parameters. S44: Based on the predicted values ​​of the four key feature parameters obtained in step S43, the predicted value of the risk index is obtained using the same method as in step S41. .

5. The intelligent measurement method for seepage of structures crossing dikes under freeze-thaw conditions according to claim 4, characterized in that, The LSTM neural network model described in step S42 adopts a two-layer LSTM unit structure, with Dropout operation introduced between layers, ReLU activation function used in the hidden layer, and linear activation function used in the output layer.

6. The intelligent measurement method for seepage of structures crossing dikes under freeze-thaw conditions according to claim 1, characterized in that, Step S5 includes: R < L 1. Under normal conditions, all parameters are within the safe threshold, and only data is recorded; L 2≤ R < L 3. Monitor status; if individual parameters show abnormalities, mark the abnormal values ​​to alert the user. L 3≤ R < L 4. Warning status: When multiple parameters show joint anomalies and the risk of LSTM prediction increases, warning SMS and emails are sent via pre-set phone numbers and email addresses. R ≥ L 5. Alarm status: If the risk index exceeds the limit or the parameter is seriously abnormal, an on-site audible and visual alarm will be triggered and an emergency message will be pushed. in, L 1. L 2. L 3. L 4. L All 5 are variable values ​​that can be determined independently based on actual circumstances, and satisfy the following conditions: L 1< L 2< L 3< L 4< L 5.

7. The intelligent measurement method for seepage of structures crossing dikes under freeze-thaw conditions according to claim 1, characterized in that, Step S6 includes: S60: Early warning mechanism and real-time risk index constructed based on step S5 R The monitoring points that have entered the warning state are selected, which are the monitoring points that may have cracks. S61: The monitoring data of the monitoring points selected in step S60 are interpolated by introducing the Kriging interpolation method, and the monitoring point data are estimated by the optimal linear unbiased estimation, which is then transformed into a three-dimensional feature map. S62: For the three-dimensional feature map generated in step S61, a three-layer 3D convolutional neural network is used to perform multi-scale feature extraction to obtain multi-channel features; S63: Construct a three-dimensional U-Net neural network model based on an encoder-decoder architecture, and use the multi-channel features obtained in step S62 as the encoder input to output the probability of crack existence and crack orientation information at the monitoring point. S64: Settings m A preliminary crack existence probability binarization is performed using a threshold as the initial judgment. Voxels with probabilities higher than the threshold are initially identified as crack regions, resulting in a preliminary crack voxel set. m These are variables that can be determined independently; S65: Apply noisy robust DBSCAN clustering to the voxel set to obtain continuous crack clusters, and perform geometric reconstruction on the continuous crack clusters. Finally, output the crack existence probability voxel map, the spatial coordinates and volume parameters of the continuous crack clusters, crack direction angle data, and a three-dimensional crack network model.

8. The intelligent measurement method for seepage of structures crossing dikes under freeze-thaw conditions according to claim 7, characterized in that, The 3D convolutional neural network model uses 3×3×3 convolutional kernels and Leaky ReLU activation function in the shallow layers; 5×5×5 convolutional kernels and ReLU activation function in the middle layers; and 7×7×7 convolutional kernels and Swish activation function in the deep layers.

9. The intelligent measurement method for leakage of structures crossing dikes under freeze-thaw conditions according to claim 7, characterized in that, The encoder of the 3D U-Net neural network model reuses and fine-tunes the pre-trained weights of the 3D convolutional neural network model; the decoder gradually upsamples through 4 layers of 3×3×3 transposed convolutional kernels, and each layer of the decoder fuses the feature maps of the encoder at the corresponding scale through skip connections; the output layer uses a 1×1×1 convolutional kernel to output two independent channels, which correspond to the probability of crack existence and crack direction information, respectively.

10. An intelligent measurement device for leakage of structures crossing dikes under freeze-thaw conditions, characterized in that, An intelligent measurement method for seepage of a structure crossing a levee under freeze-thaw conditions, applicable to any one of claims 1-9, the device comprising a soil moisture sensor (1), a displacement sensor (2), a pore water pressure sensor (3), a data acquisition and protection box (6), an intelligent power supply controller (7), and a hybrid power supply module; The soil moisture sensor (1), displacement sensor (2), and pore water pressure sensor (3) are each connected to an independent cold-resistant polyethylene sheathed cable (4). The rear half of the three cold-resistant polyethylene sheathed cables (4) are uniformly wrapped by a PVC pipe (5) and connected to the data acquisition protective box (6). The data acquisition protection box (6) is connected to an external intelligent power supply controller (7), and the intelligent power supply controller (7) is connected to the solar panel (9) and the storage battery respectively through copper core PVC sheathed flexible wire (10); The hybrid power supply module is powered by a combination of solar panels (9) and batteries (11). The batteries (11) are wrapped with a rock wool insulation layer (12), and the rock wool insulation layer (12) is covered with a shell made of corrosion-resistant and damage-resistant glass fiber reinforced plastic (13). A solar panel bracket (8) is installed at the bottom of the solar panels (9). The soil moisture sensor (1) has three sensor probes installed at the front end. The sensor is covered with a polytetrafluoroethylene composite material protective kit (15) and connected to a cold-resistant polyethylene sheathed cable (4) at the rear output port. The connection is provided with a nitrile rubber sealing ring (16). The intelligent power supply controller (7) has a display screen (20), three indicator lights (19) and an on / off button (17) installed on its panel. A wireless transmission antenna (18) is installed on the right side. A nitrile rubber sealing ring (16) is provided at the connection between the front end of the intelligent power supply controller (7) and the copper core PVC sheathed flexible wire (10). The outer shell of the data acquisition protective box (6) is an ABS engineering plastic box (21), and the inside is filled with a thick rock wool insulation layer (22). A loudspeaker (31) is embedded on the front side, and a rain shield (32) is installed on the top of the loudspeaker (31). Two heating plates (23) of the same size are installed on the rear side inside. The width of the heating plate (23) is much smaller than the width of the rear surface. There is a power supply interface (24) between the two heating plates (23). The heating plate (23) on the right side of the rear side inside is close to a DTU wireless data transmission module (28). A wireless transmission module display screen (35) is installed on the upper surface of the module. The data acquisition protective box (6) has an on / off button (29), a wireless transmission antenna (34) installed on the right side, and the direction is parallel to the side of the data acquisition protective box (6); the front surface of the DTU wireless data transmission module (28) is connected to the wireless transmission interface (27), and an RS-232 interface (26) is installed on the upper part of the wireless transmission interface (27); an LED bulb (33) is installed in the lower right corner of the data acquisition protective box (6), and is connected to the bottom surface of the data acquisition protective box (6) through the bulb base (30); a data acquisition interface (25) is installed on the left side of the data acquisition protective box (6).

Images